🧠 Cortex Memory SDK

The Smart Context Layer for Prompt Chains in LLMs

Built by Vaishakh Vipin

Overview

Cortex Memory SDK is an enterprise-grade context-aware AI system that provides intelligent memory management for Large Language Models (LLMs). It combines semantic understanding with self-evolving patterns to deliver the most relevant context for your AI applications.

🚀 Key Features

• Semantic Context Matching: Redis-backed semantic search using sentence transformers
• Self-Evolving Patterns: Advanced statistical pattern recognition for context relevance
• Multi-LLM Support: Seamless integration with Gemini, Claude, and OpenAI
• Hybrid Context Mode: Combines semantic and self-evolving context for optimal results
• Adaptive Context Selection: Automatically chooses the best context method
• Auto-Pruning System: Intelligently manages memory storage and cleanup
• Semantic Drift Detection: Monitors and adapts to changing conversation patterns

🛠️ Installation

pip install cortex-memory-sdk

📖 Quick Start

from cortex_memory import CortexClient

# Initialize the client
client = CortexClient(api_key="your_api_key")

# Generate context-aware responses
response = client.generate_with_context(
    user_id="user123",
    prompt="What did we discuss about AI yesterday?",
    provider="gemini"  # or "claude", "openai", "auto"
)

print(response)

🔧 Advanced Usage

Hybrid Context Mode

from cortex_memory.context_manager import generate_with_hybrid_context

response = generate_with_hybrid_context(
    user_id="user123",
    prompt="Explain the latest developments in AI",
    provider="claude"
)

Adaptive Context Selection

from cortex_memory.context_manager import generate_with_adaptive_context

response = generate_with_adaptive_context(
    user_id="user123",
    prompt="What are the key points from our previous meetings?",
    provider="auto"  # Automatically selects best provider
)

🏗️ Architecture

• Redis: High-performance memory storage with semantic embeddings
• Sentence Transformers: Dense vector embeddings for semantic similarity
• Statistical Pattern Recognition: Robust algorithms for context scoring
• Multi-Provider LLM Integration: Unified interface for all major LLM providers

📊 Performance

• Fast Retrieval: Redis-pipelined operations for sub-second context retrieval
• Efficient Storage: Optimized embedding storage and compression
• Scalable: Designed for enterprise-scale deployments
• Cost-Effective: Intelligent context selection reduces token usage

🔒 Security

• API key authentication
• Rate limiting and usage tracking
• Secure Redis connections
• Privacy-focused design

📚 Documentation

For detailed documentation, visit: GitHub Repository

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🆘 Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Email: vaishakh.obelisk@gmail.com

Built with ❤️ by Vaishakh Vipin

Transform your LLM applications with intelligent context management. 🧠✨